Advanced Lane Finding Project¶
The goals / steps of this project are the following:
- Compute the camera calibration matrix and distortion coefficients given a set of chessboard images.
- Apply a distortion correction to raw images.
- Use color transforms, gradients, etc., to create a thresholded binary image.
- Apply a perspective transform to rectify binary image ("birds-eye view").
- Detect lane pixels and fit to find the lane boundary.
- Determine the curvature of the lane and vehicle position with respect to center.
- Warp the detected lane boundaries back onto the original image.
- Output visual display of the lane boundaries and numerical estimation of lane curvature and vehicle position.
Step 0 : import packages¶
In [1]:
# -*- coding=UTF-8 -*-
import numpy as np
import os
import cv2
import glob
import pickle
import matplotlib.pyplot as plt
import matplotlib.image as mping
%matplotlib inline
Step 1: Compute the camera calibration matrix and distortion coefficients .¶
The camera calibration matrix was got in the Get-CameraMatrix-UndistortMatrix.ipynb file. Please run it at first to get camera matrix. Now just to load it.
In [2]:
npzfile = np.load("camera_matrix.npz")
mtx = npzfile['mtx']
dist = npzfile['dist']
Examples of undistorted image¶
In [3]:
def imgUndistort(img, mtx=mtx, dist=dist):
"""
Undistort image
Arguments:
img: source image
mtx: camera internal matrix
dist: distortion coefficients
"""
return cv2.undistort(img, mtx, dist, None, mtx)
def show_undistort_image(img, mtx, dist, bBGR=False):
"""
Comparely show origin image and undistort image
Arguments:
img: source image
mtx: camera internal matrix
dist: distortion coefficients
bBGR: bool value, if BGR channel order
"""
imgRGB = img
if bBGR:
imgRGB = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
img_undistort = imgUndistort(imgRGB, mtx, dist)
plt.figure(figsize=(16,8))
plt.subplot(1,2,1)
plt.title("original Image", fontsize=16)
plt.imshow(imgRGB)
plt.subplot(1,2,2)
plt.title("Undistorted Image", fontsize=16)
plt.imshow(img_undistort)
def showImages(images, label=None, cols=3, figsize=(14,14), ticksshow=True):
"""
Show cols images per colum
Arguments:
images: image list or array-like
label: image label, list
cols: the number of images per colums
ticksshow: whether show ticks, bool value
"""
rows = (len(images) + cols - 1)//cols
plt.figure(figsize=figsize)
for i, image in enumerate(images):
plt.subplot(rows, cols, i+1)
# use gray scale color map if there is only one channel
imgShape = image.shape
cmap = None
if len(imgShape) == 2:
cmap = "gray"
elif imgShape[2] == 1:
image = image[:,:,0]
cmap = "gray"
plt.imshow(image, cmap=cmap)
if label != None and label[i] != None:
plt.title(label[i], fontsize=12)
if not ticksshow:
plt.xticks([])
plt.yticks([])
plt.tight_layout(pad=0, h_pad=0, w_pad=0)
plt.show()
In [4]:
show_undistort_image(cv2.imread("camera_cal/calibration2.jpg"),mtx,dist)
Step 2: Apply a distortion correction to raw images.¶
In [5]:
testImageList = glob.glob('test_images/*.jpg')
testImages = [cv2.cvtColor(cv2.imread(img), cv2.COLOR_BGR2RGB) for img in testImageList]
images_undistort = list(map(imgUndistort, testImages))
In [6]:
# for filename in testImageList:
# img = cv2.imread(filename)
# imgRGB = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# show_undistort_image(imgRGB, mtx, dist)
showImages(images_undistort, testImageList, cols=3, figsize=(14,14), ticksshow=False)
In [7]:
def splitChannels(img):
"""
Split image into three channels in HSV, HLS and Lab separately
Arguments:
img: source image
"""
channel_images=[]
channel_images_label=[]
#origin RGB image
channel_images.append(img)
channel_images_label.append('Original Image')
gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
channel_images.append(gray)
channel_images_label.append('Gray Image')
grayequalizeHist = cv2.equalizeHist(gray)
channel_images.append(grayequalizeHist)
channel_images_label.append('EqualizeHist Gray Image')
#HLS image
hls = cv2.cvtColor(img, cv2.COLOR_RGB2HLS)
h = hls[:,:,0]
l = hls[:,:,1]
s = hls[:,:,2]
channel_images.append(h)
channel_images_label.append('HLS: H Channel Image')
channel_images.append(l)
channel_images_label.append('HLS: L Channel Image')
channel_images.append(s)
channel_images_label.append('HLS: S Channel Image')
#HSV image
hsv = cv2.cvtColor(img, cv2.COLOR_RGB2HSV)
h = hsv[:,:,0]
s = hsv[:,:,1]
v = hsv[:,:,2]
channel_images.append(h)
channel_images_label.append('HSV: H Channel Image')
channel_images.append(s)
channel_images_label.append('HSV: S Channel Image')
channel_images.append(v)
channel_images_label.append('HSV: V Channel Image')
#Lab image
lab = cv2.cvtColor(img, cv2.COLOR_RGB2Lab)
l = lab[:,:,0]
a = lab[:,:,1]
b = lab[:,:,2]
channel_images.append(l)
channel_images_label.append('Lab: L Channel Image')
channel_images.append(a)
channel_images_label.append('Lab: a Channel Image')
channel_images.append(b)
channel_images_label.append('Lab: b Channel Image')
return channel_images,channel_images_label
In [8]:
testImgRGB = cv2.cvtColor(cv2.imread(testImageList[0]), cv2.COLOR_BGR2RGB)
testImg_undistort = cv2.undistort(testImgRGB, mtx, dist, None, mtx)
images, label = splitChannels(testImg_undistort)
showImages(images, label, figsize=(14,14), ticksshow=False)
Define color thresh functions¶
In [9]:
def color_thresh(img, threshLow, threshHigh, colorSpace="HSV", oneChannel=None):
"""Convert color space to another, return binary image
Arguments:
img: RGB channel order
colorSpace: "RGB", "HSV", "HSL"
threshLow: if oneChannel=None (channel1, channel2, channel3) tuple, else single value
threshHigh: if oneChannel=None (channel1, channel2, channel3) tuple, else single value
oneChannel: appoint one channel
"""
dstImg = np.zeros_like(img)
if colorSpace == "HSV":
dstImg = cv2.cvtColor(img, cv2.COLOR_RGB2HSV)
if colorSpace == "HSL":
dstImg = cv2.cvtColor(img, cv2.COLOR_RGB2HLS)
if colorSpace == "Lab":
dstImg = cv2.cvtColor(img, cv2.COLOR_RGB2Lab)
if colorSpace == "RGB":
dstImg = img
# Return color select mask
binary_output = np.zeros((img.shape[0], img.shape[1]))
if oneChannel:
oneChannelImage = dstImg[:,:,oneChannel]
if colorSpace == "HSV":
oneChannelImage = cv2.equalizeHist(oneChannelImage)
binary_output[(oneChannelImage >= threshLow) & (oneChannelImage <= threshHigh)] = 1
else:
binary_output[(dstImg[:,:,0] >= threshLow[0]) & (dstImg[:,:,0] <= threshHigh[0]) & (dstImg[:,:,1] >= threshLow[1]) & (dstImg[:,:,1] <= threshHigh[1]) & (dstImg[:,:,2] >= threshLow[2]) & (dstImg[:,:,2] <= threshHigh[2])] = 1
return binary_output
Yellow Lane filtered by Lab color space in b Channel¶
In [10]:
color_binary = list(map(lambda img: color_thresh(img, 150, 255, colorSpace="Lab", oneChannel=2), images_undistort))
showImages(color_binary, testImageList, ticksshow=False)
White Lane filtered by HSL color space in Lightness Channel¶
In [11]:
color_binary = list(map(lambda img: color_thresh(img, 210, 255, colorSpace="HSL", oneChannel=1), images_undistort))
showImages(color_binary, testImageList, ticksshow=False)
Road Area Selected by HSV color space in Value Channel¶
In [12]:
color_binary = list(map(lambda img: color_thresh(img, 145, 255, colorSpace="HSV", oneChannel=2), images_undistort))
showImages(color_binary, testImageList, ticksshow=False)
Define Gradient and sobel functions¶
In [13]:
def sobel_x(img, sobel_kernel=3, min_thresh = 20, max_thresh =100, colorSpace="HSL"):
"""
Filter out horizontal noise, return binary image
Arguments:
img: source image
colorSpace: "RGB", "HSL", "HSV"
sobel_kernel: kernel size
min_thes: min thresh
max_thes: max thresh
"""
dstImg = np.zeros_like(img)
if colorSpace == "HSV":
dstImg = cv2.cvtColor(img, cv2.COLOR_RGB2HSV)
if colorSpace == "HSL":
dstImg = cv2.cvtColor(img, cv2.COLOR_RGB2HLS)
if colorSpace == "RGB":
dstImg = img
#Channels L and S from HLS
sobelx1 = cv2.Sobel(dstImg[:,:,1], cv2.CV_64F, 1, 0, ksize=sobel_kernel)
sobelx2 = cv2.Sobel(dstImg[:,:,2], cv2.CV_64F, 1, 0, ksize=sobel_kernel)
# Scale to 8-bit (0 - 255) and convert to type = np.uint8
scaled_sobelx1 = np.uint8(255*sobelx1/ np.max(sobelx1))
scaled_sobelx2 = np.uint8(255*sobelx2/ np.max(sobelx2))
# Create a binary mask where thresholds are met
binary_outputx1 = np.zeros_like(scaled_sobelx1)
binary_outputx1[(scaled_sobelx1 >= min_thresh) & (scaled_sobelx1 <= max_thresh)] = 1
binary_outputx2 = np.zeros_like(scaled_sobelx2)
binary_outputx2[(scaled_sobelx2 >= min_thresh) & (scaled_sobelx2 <= max_thresh)] = 1
binary_output = np.zeros_like(scaled_sobelx1)
binary_output[(binary_outputx1 ==1) | (binary_outputx2 ==1)]=1
return binary_output
def sobel_thresh(img, orientation='x', sobel_kernel=5, min_thresh=20, max_thresh=255):
gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
gray = cv2.equalizeHist(gray)
if orientation == 'x':
abs_sobel = np.absolute(cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize=sobel_kernel))
if orientation == 'y':
abs_sobel = np.absolute(cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize=sobel_kernel))
scaled_sobel = np.uint8(255*abs_sobel/ np.max(abs_sobel))
binary_output = np.zeros_like(scaled_sobel)
binary_output[(scaled_sobel >= min_thresh) & (scaled_sobel <= max_thresh)] = 1
return binary_output
def gradMag_thresh(img, sobel_kernel=3, gradmag_thresh=(50, 255)):
"""
Calulate magnitude of gradient, return binary image
Arguments:
img: source image
sobel_kernel: kernel size
gradmag_thresh: magnitude of gradient threshold
"""
# Convert to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
# Take the gradient in x and y separately
sobelx = cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize=sobel_kernel)
sobely = cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize=sobel_kernel)
# Calculate the magnitude of gradient
gradmag = np.sqrt(sobelx**2 + sobely**2)
# Scale to 8-bit (0 - 255) and convert to type = np.uint8
scaled_sobel = np.uint8(255*gradmag / np.max(gradmag))
# Create a binary mask where mag thresholds are met
binary_output = np.zeros_like(scaled_sobel)
binary_output[(scaled_sobel >= gradmag_thresh[0]) & (scaled_sobel <= gradmag_thresh[1])] = 1
return binary_output
#Direction threshold
def direction_thresh(img, sobel_kernel=3, thresh=(0.7, np.pi/2)):
"""
Using angle thesh
Arguments:
img: source image
sobel_kernel: kernel size
thresh: angle rangement
"""
# Convert to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
# Take the gradient in x and y separately
sobelx = cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize=sobel_kernel)
sobely = cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize=sobel_kernel)
# Take the absolute value of the x and y gradients
abs_sobelx = np.absolute(sobelx)
abs_sobely = np.absolute(sobely)
# Use np.arctan2(abs_sobely, abs_sobelx) to calculate the direction of the gradient
absgraddirection = np.arctan2(abs_sobely, abs_sobelx)
# Create a binary mask where direction thresholds are met
binary_output = np.zeros_like(absgraddirection)
binary_output[(absgraddirection >= thresh[0]) & (absgraddirection <= thresh[1])] = 1
return binary_output
#Both Magnitude and direction threshold
def gradMag_direction_thresh(img, sobel_kernel=3, mag_thresh=(50, 255), dir_thresh=(0.7, np.pi/2)):
"""
Both using Magnitude of gradient and angle direction thresh, return binary image
Arguments:
img: source image
sobel_kernel: kernel size
mag_thresh: magnitude of gradient thresh
dir_thresh: direction thresh
"""
# Convert to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
# Take the gradient in x and y separately
sobelx = cv2.Sobel(img, cv2.CV_64F, 1, 0, ksize=sobel_kernel)
sobely = cv2.Sobel(img, cv2.CV_64F, 0, 1, ksize=sobel_kernel)
# Calculate the magnitude of gradient
gradmag = np.sqrt(sobelx**2 + sobely**2)
#Calculate angle
abs_sobelx = np.absolute(sobelx)
abs_sobely = np.absolute(sobely)
absgraddir = np.arctan2(abs_sobely, abs_sobelx)
# Scale to 8-bit (0 - 255) and convert to type = np.uint8
scaled_sobel = np.uint8(255*gradmag / np.max(gradmag))
# Create a binary mask where mag thresholds are met
binary_output = np.zeros_like(scaled_sobel)
binary_output[(scaled_sobel >= mag_thresh[0]) & (scaled_sobel <= mag_thresh[1]) & (absgraddir >= dir_thresh[0]) & (absgraddir <= dir_thresh[1]) ] = 1
return binary_output
Examples of gradient thresholds¶
In [14]:
gradient_binary = list(map(sobel_x, images_undistort))
showImages(gradient_binary, testImageList, ticksshow=False)
In [15]:
gradient_binary = list(map(sobel_thresh, images_undistort))
showImages(gradient_binary, testImageList, ticksshow=False)
In [16]:
gradient_binary = list(map(gradMag_thresh, images_undistort))
showImages(gradient_binary, testImageList, ticksshow=False)
In [17]:
def pipeline(img):
sobel_thresh_binary = sobel_thresh(img)
hsv_thresh_binary = color_thresh(img, 145, 255, colorSpace="HSV", oneChannel=2)
hsl_thresh_binary = color_thresh(img, 210, 255, colorSpace="HSL", oneChannel=1)
lab_thresh_binary = color_thresh(img, 150, 255, colorSpace="Lab", oneChannel=2)
combined_binary = np.zeros_like(sobel_thresh_binary)
combined_binary[ ((sobel_thresh_binary == 1) & (hsv_thresh_binary == 1)) |(hsl_thresh_binary == 1)| (lab_thresh_binary == 1)] = 1
return combined_binary
In [18]:
filted_binary = list(map(pipeline, images_undistort))
showImages(filted_binary, testImageList, ticksshow=False)
In [19]:
def perspective_transform_matrix(img, show_debug=False):
img_size = img.shape[:2][::-1]
src = np.float32(\
[[(img_size[0] / 2) - 75, img_size[1] / 2 + 110],\
[((img_size[0] / 6) - 10), img_size[1]],\
[(img_size[0] * 5 / 6) + 60, img_size[1]],\
[(img_size[0] / 2 + 75), img_size[1] / 2 + 110]])
dst = np.float32(\
[[(img_size[0] / 4), 0],\
[(img_size[0] / 4), img_size[1]],\
[(img_size[0] * 3 / 4), img_size[1]],\
[(img_size[0] * 3 / 4), 0]])
M = cv2.getPerspectiveTransform(src, dst)
Minv = cv2.getPerspectiveTransform(dst, src)
if show_debug == True:
print("source view point")
print(src)
print("Transformed view point")
print(dst)
plt.figure(figsize=(16, 16))
plt.subplot(1, 2, 1)
plt.imshow(img)
plt.plot(src[:,0], src[:,1], 'rx')
plt.title('Original Image')
plt.subplot(1, 2, 2)
warped = cv2.warpPerspective(img, M, img_size, flags=cv2.INTER_LINEAR)
plt.imshow(warped)
plt.plot(dst[:,0], dst[:,1], 'rx')
plt.title('Perspective Transformed Image')
return M, Minv
In [20]:
straightimage = mping.imread('test_images/straight_lines1.jpg')
M,Minv = perspective_transform_matrix(straightimage,show_debug=True)
In [21]:
def warpImage(img):
img_size = img.shape[:2][::-1]
M, Minv = perspective_transform_matrix(img)
return cv2.warpPerspective(img, M, img_size, flags=cv2.INTER_LINEAR)
Lane marked Perspective Transform¶
In [22]:
warped_images = list(map(warpImage, filted_binary))
showImages(warped_images, testImageList, ticksshow=False)
In [23]:
def show_histogram(binary_image):
histogram = np.sum(binary_image[binary_image.shape[0]//2:, :], axis=0)
plt.plot(histogram)
In [24]:
show_histogram(warped_images[0])
In [25]:
def fitlines(binary_warped, nwindows=9, margin=100, minpix=50, bias=0.2):
"""
margin: Set the width of the windows +/- margin
minpix: Set minimum number of pixels found to recenter window
bias: left start and right start offest from middpoint
"""
# Assuming you have created a warped binary image called "binary_warped"
# Take a histogram of the bottom half of the image
histogram = np.sum(binary_warped[binary_warped.shape[0]//2:,:], axis=0)
# Create an output image to draw on and visualize the result
out_img = np.dstack((binary_warped, binary_warped, binary_warped))*255
# Find the peak of the left and right halves of the histogram
# These will be the starting point for the left and right lines
midpoint = np.int(histogram.shape[0]/2)
leftstart = int((1-bias)*midpoint)
rightstart = int((1+bias)*midpoint)
leftx_base = np.argmax(histogram[:leftstart])
rightx_base = np.argmax(histogram[rightstart:]) + rightstart
# Set height of windows
window_height = np.int(binary_warped.shape[0]/nwindows)
# Identify the x and y positions of all nonzero pixels in the image
nonzero = binary_warped.nonzero()
nonzeroy = np.array(nonzero[0])
nonzerox = np.array(nonzero[1])
# Current positions to be updated for each window
leftx_current = leftx_base
rightx_current = rightx_base
left_lane_inds = []
right_lane_inds = []
# Step through the windows one by one
for window in range(nwindows):
# Identify window boundaries in x and y (and right and left)
win_y_low = binary_warped.shape[0] - (window+1)*window_height
win_y_high = binary_warped.shape[0] - window*window_height
win_xleft_low = leftx_current - margin
win_xleft_high = leftx_current + margin
win_xright_low = rightx_current - margin
win_xright_high = rightx_current + margin
# Draw the windows on the visualization image
cv2.rectangle(out_img,(win_xleft_low,win_y_low),(win_xleft_high,win_y_high),(0,255,0), 4)
cv2.rectangle(out_img,(win_xright_low,win_y_low),(win_xright_high,win_y_high),(0,255,0), 4)
# Identify the nonzero pixels in x and y within the window
good_left_inds = ((nonzeroy >= win_y_low) & (nonzeroy < win_y_high) & (nonzerox >= win_xleft_low) & (nonzerox < win_xleft_high)).nonzero()[0]
good_right_inds = ((nonzeroy >= win_y_low) & (nonzeroy < win_y_high) & (nonzerox >= win_xright_low) & (nonzerox < win_xright_high)).nonzero()[0]
# Append these indices to the lists
left_lane_inds.append(good_left_inds)
right_lane_inds.append(good_right_inds)
# If you found > minpix pixels, recenter next window on their mean position
if len(good_left_inds) > minpix:
leftx_current = np.int(np.mean(nonzerox[good_left_inds]))
if len(good_right_inds) > minpix:
rightx_current = np.int(np.mean(nonzerox[good_right_inds]))
# Concatenate the arrays of indices
left_lane_inds = np.concatenate(left_lane_inds)
right_lane_inds = np.concatenate(right_lane_inds)
# Extract left and right line pixel positions
leftx = nonzerox[left_lane_inds]
lefty = nonzeroy[left_lane_inds]
rightx = nonzerox[right_lane_inds]
righty = nonzeroy[right_lane_inds]
out_img[nonzeroy[left_lane_inds], nonzerox[left_lane_inds]] = [255, 0, 0]
out_img[nonzeroy[right_lane_inds], nonzerox[right_lane_inds]] = [0, 0, 255]
left_fit = None
right_fit = None
ploty = np.linspace(0, binary_warped.shape[0]-1, binary_warped.shape[0] )
# Fit a second order polynomial to each
if len(leftx) != 0:
left_fit = np.polyfit(lefty, leftx, 2)
p_left = np.poly1d(left_fit)
left_fitx = p_left(ploty)
left_line_pts = np.array([np.transpose(np.vstack([left_fitx, ploty]))])
cv2.polylines(out_img, np.int_(left_line_pts), isClosed=False, color=(255,255,0), thickness=5)
if len(rightx) != 0:
right_fit = np.polyfit(righty, rightx, 2)
p_right = np.poly1d(right_fit)
right_fitx = p_right(ploty)
right_line_pts = np.array([np.transpose(np.vstack([right_fitx, ploty]))])
cv2.polylines(out_img, np.int_(right_line_pts), isClosed=False, color=(255,255,0), thickness=5)
return out_img, left_fit, right_fit
In [26]:
fitline_images = list(map(fitlines, warped_images))
fittedimges = [s[0] for s in fitline_images]
left_fits = [s[1] for s in fitline_images]
right_fits = [s[2] for s in fitline_images]
showImages(fittedimges, testImageList, ticksshow=False)
In [27]:
def fitted_lanes(binary_warped, left_fit, right_fit, margin = 100, minpix = 50):
# Assume you now have a new warped binary image
# from the next frame of video (also called "binary_warped")
# It's now much easier to find line pixels!
nonzero = binary_warped.nonzero()
nonzeroy = np.array(nonzero[0])
nonzerox = np.array(nonzero[1])
left_lane_inds = ((nonzerox > (left_fit[0]*(nonzeroy**2) + left_fit[1]*nonzeroy +
left_fit[2] - margin)) & (nonzerox < (left_fit[0]*(nonzeroy**2) +
left_fit[1]*nonzeroy + left_fit[2] + margin)))
right_lane_inds = ((nonzerox > (right_fit[0]*(nonzeroy**2) + right_fit[1]*nonzeroy +
right_fit[2] - margin)) & (nonzerox < (right_fit[0]*(nonzeroy**2) +
right_fit[1]*nonzeroy + right_fit[2] + margin)))
# Again, extract left and right line pixel positions
leftx = nonzerox[left_lane_inds]
lefty = nonzeroy[left_lane_inds]
rightx = nonzerox[right_lane_inds]
righty = nonzeroy[right_lane_inds]
# Create an image to draw on and an image to show the selection window
out_img = np.dstack((binary_warped, binary_warped, binary_warped))*255
window_img = np.zeros_like(out_img)
# Color in left and right line pixels
out_img[nonzeroy[left_lane_inds], nonzerox[left_lane_inds]] = [255, 0, 0]
out_img[nonzeroy[right_lane_inds], nonzerox[right_lane_inds]] = [0, 0, 255]
left_new_fit = None
right_new_fit = None
ploty = np.linspace(0, binary_warped.shape[0]-1, binary_warped.shape[0] )
# Fit a second order polynomial to each
p_left = np.poly1d(left_fit)
left_fitx = p_left(ploty)
left_line_window1 = np.array([np.transpose(np.vstack([left_fitx-margin, ploty]))])
left_line_window2 = np.array([np.flipud(np.transpose(np.vstack([left_fitx+margin,
ploty])))])
left_line_pts = np.hstack((left_line_window1, left_line_window2))
cv2.fillPoly(window_img, np.int_([left_line_pts]), (0,255, 0))
p_right = np.poly1d(right_fit)
right_fitx = p_right(ploty)
# Generate a polygon to illustrate the search window area
right_line_window1 = np.array([np.transpose(np.vstack([right_fitx-margin, ploty]))])
right_line_window2 = np.array([np.flipud(np.transpose(np.vstack([right_fitx+margin,
ploty])))])
right_line_pts = np.hstack((right_line_window1, right_line_window2))
# Draw the lane onto the warped blank image
cv2.fillPoly(window_img, np.int_([right_line_pts]), (0,255, 0))
if len(leftx) != 0:
left_new_fit = np.polyfit(lefty, leftx, 2)
p_left = np.poly1d(left_new_fit)
left_fitx = p_left(ploty)
# Generate a polygon to illustrate the search window area
# And recast the x and y points into usable format for cv2.fillPoly()
left_line_pts = np.array([np.transpose(np.vstack([left_fitx, ploty]))])
cv2.polylines(out_img, np.int_(left_line_pts), isClosed=False, color=(255,255,0), thickness=5)
if len(rightx) != 0:
right_new_fit = np.polyfit(righty, rightx, 2)
p_right = np.poly1d(right_new_fit)
right_fitx = p_right(ploty)
right_line_pts = np.array([np.transpose(np.vstack([right_fitx, ploty]))])
cv2.polylines(out_img, np.int_(right_line_pts), isClosed=False, color=(255,255,0), thickness=5)
result = cv2.addWeighted(out_img, 1, window_img, 0.3, 0)
return result,left_new_fit,right_new_fit
In [28]:
recogin_images = list(map(fitted_lanes, warped_images, left_fits, right_fits))
fittedimges = [s[0] for s in recogin_images]
left_fits = [s[1] for s in recogin_images]
right_fits = [s[2] for s in recogin_images]
showImages(fittedimges, testImageList, ticksshow=False)
In [29]:
def calculate_curvature(binary_warped, line_fit):
# Define conversions in x and y from pixels space to meters
ym_per_pix = 30/720 # meters per pixel in y dimension
xm_per_pix = 3.7/700 # meters per pixel in x dimension
ploty = np.linspace(0, binary_warped.shape[0]-1, binary_warped.shape[0] )
line_fitx = line_fit[0]*ploty**2 + line_fit[1]*ploty + line_fit[2]
y_eval = np.max(ploty)
# Fit new polynomials to x,y in world space
line_fit_cr = np.polyfit(ploty*ym_per_pix, line_fitx*xm_per_pix, 2)
# Calculate the new radii of curvature
cal_curverad = ((1 + (2*line_fit_cr[0]*y_eval*ym_per_pix + line_fit_cr[1])**2)**1.5) / np.absolute(2*line_fit_cr[0])
# Calculate line position offset to center
p_line = np.poly1d(line_fit)
line_fitx_1 = p_line(y_eval)
center_offset = (line_fitx_1 - binary_warped.shape[1]/2)*xm_per_pix
return cal_curverad, center_offset
Setp 7: Warp the detected lane boundaries back onto the original image.¶
In [30]:
def unwarp(img, left_fit, right_fit):
img_size = img.shape[:2][::-1]
M, Minv = perspective_transform_matrix(img)
if left_fit.any() != None and right_fit.any() != None:
color_warp = np.zeros_like(img).astype(np.uint8)
ploty = np.linspace(0, img.shape[0]-1, img.shape[0] )
left_fitx = left_fit[0]*ploty**2 + left_fit[1]*ploty + left_fit[2]
right_fitx = right_fit[0]*ploty**2 + right_fit[1]*ploty + right_fit[2]
# Recast the x and y points into usable format for cv2.fillPoly()
pts_left = np.array([np.transpose(np.vstack([left_fitx, ploty]))])
pts_right = np.array([np.flipud(np.transpose(np.vstack([right_fitx, ploty])))])
pts = np.hstack((pts_left, pts_right))
# Draw the lane onto the warped blank image
cv2.fillPoly(color_warp, np.int_([pts]), (0,255, 0))
# Warp the blank back to original image space using inverse perspective matrix (Minv)
newwarp = cv2.warpPerspective(color_warp, Minv, (img.shape[1], img.shape[0]))
# Combine the result with the original image
result = cv2.addWeighted(img, 1, newwarp, 0.3, 0)
else:
result = img
return result
In [31]:
unwarp_images = list(map(unwarp, images_undistort, left_fits, right_fits))
showImages(unwarp_images, testImageList)
Setp 8: Output visual display of the lane boundaries and numerical estimation of lane curvature and vehicle position.¶
In [32]:
def draw_estimation(img, left_fit, right_fit):
if left_fit.any() != None and right_fit.any() != None:
left_cur, left_center_offest = calculate_curvature(img, left_fit)
right_cur, right_center_offest = calculate_curvature(img, right_fit)
center_offest = (left_center_offest + right_center_offest)/2.0
cv2.putText(img,'left curvature:{:.2f}m'.format(left_cur),(10,60),cv2.FONT_HERSHEY_COMPLEX,2,(255,255,255),3)
cv2.putText(img,'right curvature:{:.2f}m'.format(right_cur),(10,130),cv2.FONT_HERSHEY_COMPLEX,2,(255,255,255),3)
cv2.putText(img,'center offest:{:.2f}m'.format(center_offest),(10,200),cv2.FONT_HERSHEY_COMPLEX,2,(255,255,255),3)
return img
In [33]:
res_images= list(map(draw_estimation, unwarp_images, left_fits, right_fits))
showImages(res_images,testImageList)
In [34]:
from collections import deque
# Define a class to receive the characteristics of each line detection
class Line():
def __init__(self, maxlen=10):
# was the line detected in the last iteration?
self.detected = False
# x values of the last n fits of the line
self.recent_xfitted = deque(maxlen = maxlen)
#average x values of the fitted line over the last n iterations
self.bestx = None
#polynomial coefficients averaged over the last n iterations
self.best_fit = None
#polynomial coefficients for the most recent fit
self.current_fit = [np.array([False])]
#radius of curvature of the line in some units
self.radius_of_curvature = None
#distance in meters of vehicle center from the line
self.line_base_pos = None
#difference in fit coefficients between last and new fits
self.diffs = np.array([0,0,0], dtype='float')
def update_detected(self, img, detected=True):
self.detected = detected
#if lane lines do not dectect correctly,remove from deque
if self.detected == False and len(self.recent_xfitted) > 0:
self.recent_xfitted.pop()
self.update_bestx(img)
self.update_best_fit(img)
def update_recent_xfitted(self,img):
yvals = np.linspace(0,img.shape[0]-1,img.shape[0])
current_xval = self.current_fit[0]*yvals**2 + self.current_fit[1]*yvals + self.current_fit[2]
self.recent_xfitted.append(current_xval)
def update_bestx(self,img):
if len(self.recent_xfitted) > 0:
self.bestx = np.mean(self.recent_xfitted, axis=0)
else:
self.bestx = None
def update_best_fit(self,img):
if self.bestx.any() != None:
yvals = np.linspace(0, img.shape[0]-1, img.shape[0] )
line_new_fit = np.polyfit(yvals, self.bestx, 2)
self.best_fit = line_new_fit
else:
self.best_fit = None
def update_current_fit(self, fit):
self.current_fit = fit
def update_radius_of_curvature(self, img):
self.radius_of_curvature, _ = calculate_curvature(img,self.current_fit)
def update_line_base_pos(self, img):
_, self.line_base_pos = calculate_curvature(img,self.current_fit)
def update_diffs(self):
if np.any(self.best_fit) != None:
self.diffs = self.current_fit - self.best_fit
else:
self.diffs = np.array([0,0,0], dtype='float')
def upadate_data(self,img,fit):
self.update_current_fit(fit)
self.update_recent_xfitted(img)
self.update_diffs()
self.update_radius_of_curvature(img)
self.update_line_base_pos(img)
def reset_data(self):
self.recent_xfitted.clear()
self.bestx = None
self.best_fit = None
In [35]:
class Linedectect():
def __init__(self,maxlen=15):
# was the line detected in the last iteration?
self.detected = False
self.maxlen = maxlen
self.lanedeparturecount = 0
self.framenumbercnt = 0
# Define left and right lines for detection
self.left_lines = Line(self.maxlen)
self.right_lines = Line(self.maxlen)
def sanity_check(self,left_line,right_line):
curv_ratio_threshold = 25
distance_min_diff_bias = 200 #pixel in x dimension
distance_pos_max_diff_bias = 1000 #pixel in x dimension
distance_deviation_threshold = 100 #pixel in x dimension
#Checking that they have similiar curvature
left_radius_of_curvature = left_line.radius_of_curvature
right_radius_of_curvature = right_line.radius_of_curvature
curvature_ratio = left_radius_of_curvature/right_radius_of_curvature
if curvature_ratio > curv_ratio_threshold or curvature_ratio <1.0/curv_ratio_threshold:
# print("curvature_ratio:")
# print(left_radius_of_curvature)
# print(right_radius_of_curvature)
# print(curvature_ratio)
return False
#Checking that they are seperated by approximately the right distance horizontally
left_fitx = left_line.recent_xfitted[-1]
right_fitx = right_line.recent_xfitted[-1]
lane_horizon_distance = right_fitx - left_fitx
lane_distance_max = max(lane_horizon_distance)
lane_distance_min = min(lane_horizon_distance)
if lane_distance_min < distance_min_diff_bias or lane_distance_max > distance_pos_max_diff_bias:
# print("lane_distance:")
# print(lane_distance_max)
# print(lane_distance_min)
return False
# Checking that they are roughly parallel
distance_deviation = np.std(lane_horizon_distance)
if distance_deviation >= distance_deviation_threshold:
# print("distance_deviation:")
# print(distance_deviation)
return False
left_line_diff = np.absolute(left_line.diffs)
if left_line_diff[0] > 0.01 or left_line_diff[1] > 0.5 or left_line_diff[2] > 100:
# print("left_line_diff:")
# print(left_line_diff)
return False
right_line_diff = np.absolute(right_line.diffs)
if right_line_diff[0] > 0.01 or right_line_diff[1] > 0.5 or right_line_diff[2] > 100:
# print("right_line_diff:")
# print(right_line_diff)
return False
left_right_line_diff = np.absolute(left_line.current_fit - right_line.current_fit)
if left_right_line_diff[0] > 0.01 or left_right_line_diff[1] > 0.5 :
# print("left_right_line_diff:")
return False
return True
def line_find_process(self, image, left_line, right_line):
if self.lanedeparturecount >= self.maxlen:
left_line.reset_data()
right_line.reset_data()
if np.any(left_line.best_fit) != None and np.any(right_line.best_fit) != None:
fittedimg,recent_leftfitted,recent_rightfitted = fitted_lanes(image,left_line.best_fit,right_line.best_fit)
else:
fittedimg,recent_leftfitted,recent_rightfitted = fitlines(image)
if recent_leftfitted.any() != None and recent_rightfitted.any() != None:
left_line.upadate_data(image,recent_leftfitted)
right_line.upadate_data(image,recent_rightfitted)
if self.sanity_check(left_line,right_line) == True:
left_line.update_detected(image,True)
right_line.update_detected(image,True)
self.lanedeparturecount = 0
else:
left_line.update_detected(image,False)
right_line.update_detected(image,False)
self.lanedeparturecount += 1
else:
self.lanedeparturecount += 1
return fittedimg,left_line.best_fit,right_line.best_fit
def image_lane_find_process(self,image,debugcombined = True,framenumber = None):
undistort_images = imgUndistort(image)
pipe_images = pipeline(undistort_images)
warp_images = warpImage(pipe_images)
fittedimg,recent_leftfitted,recent_rightfitted = self.line_find_process(warp_images,self.left_lines,self.right_lines)
unwarp_images = unwarp(undistort_images,recent_leftfitted,recent_rightfitted)
res_images = draw_estimation(unwarp_images,recent_leftfitted,recent_rightfitted)
if debugcombined == True:
# Calculate the size of screens
result_screen_w = unwarp_images.shape[1]
result_screen_h = unwarp_images.shape[0]
debug_screen_w = np.int(result_screen_w/3)
debug_screen_h = np.int(result_screen_h/3)
screen_w = result_screen_w + debug_screen_w
screen_h = result_screen_h
# Assign result image to the screen
#show screen
screen = np.zeros((screen_h,screen_w,3),dtype=np.uint8)
if framenumber != None:
cv2.putText(unwarp_images,'frame index:{:}'.format(framenumber),(10,270),cv2.FONT_HERSHEY_COMPLEX,2,(255,255,255),3)
cv2.putText(unwarp_images,'error count:{:}'.format(self.lanedeparturecount),(10,340),cv2.FONT_HERSHEY_COMPLEX,2,(255,255,255),3)
screen[0:result_screen_h,0:result_screen_w] = res_images
screen[0:result_screen_h,result_screen_w-2:result_screen_w] = (50,255,255)
# Assign debug image to the screen
debug_img_1 = np.dstack((pipe_images,pipe_images,pipe_images))*255
screen[0:debug_screen_h,result_screen_w:,:] = cv2.resize(debug_img_1,(debug_screen_w,debug_screen_h))
screen[debug_screen_h-2:debug_screen_h,result_screen_w:,:] = (50,255,255)
debug_img_2 = np.dstack((warp_images,warp_images,warp_images))*255
screen[debug_screen_h : debug_screen_h*2,result_screen_w:,:] =cv2.resize(debug_img_2,(debug_screen_w,debug_screen_h))
screen[debug_screen_h*2-2 : debug_screen_h*2,result_screen_w:,:] = (50,255,255)
debug_img_3 = fittedimg
screen[debug_screen_h*2 : debug_screen_h*3,result_screen_w:,:] =cv2.resize(debug_img_3,(debug_screen_w,debug_screen_h))
return screen
else:
return unwarp_images
def video_process_show(self,image):
res_img = self.image_lane_find_process(image,framenumber=self.framenumbercnt)
self.framenumbercnt += 1
return res_img
In [36]:
image_process_test_image = testImages[0]
plt.figure(figsize=(8, 8))
plt.imshow(image_process_test_image)
plt.title('Original Image')
plt.figure(figsize=(16, 16))
L = Linedectect()
warp_image_process_test_image = L.video_process_show(image_process_test_image)
plt.imshow(warp_image_process_test_image)
plt.title('Result Image')
Out[36]:
Test Videos¶
In [37]:
from moviepy.editor import VideoFileClip
from IPython.display import HTML
In [38]:
project_source = "test_videos/project_video.mp4"
project_output = "output_videos/project_video_output.mp4"
## To speed up the testing process you may want to try your pipeline on a shorter subclip of the video
## To do so add .subclip(start_second,end_second) to the end of the line below
## Where start_second and end_second are integer values representing the start and end of the subclip
## You may also uncomment the following line for a subclip of the first 5 seconds
##clip1 = VideoFileClip("test_videos/solidWhiteRight.mp4").subclip(0,5)
L = Linedectect()
clip1 = VideoFileClip(project_source)
line_clip = clip1.fl_image(L.video_process_show) #NOTE: this function expects color images!!
%time line_clip.write_videofile(project_output, audio=False)
In [39]:
HTML("""
<video width="960" height="450" controls>
<source src="{0}">
</video>
""".format(project_output))
Out[39]: